Visible photoelectrochemical water splitting into H₂ and O₂ in a dye-sensitized photoelectrosynthesis cell

Significance

Mesoporous SnO₂/TiO₂ core/shell nanostructured electrodes derivatized with a surface-bound Ru(II) polypyridyl-based chromophore–catalyst assembly are used for water splitting into H₂ and O₂ with visible light in a dye-sensitized photoelectrosynthesis cell. Photocurrents with a small applied bias are among the highest reported. Stabilization of the assembly on the surface of the TiO₂ shell by using atomic layer deposition to deposit overlayers of Al₂O₃ or TiO₂ results in long-term water splitting even in a phosphate buffer at pH 7.

Abstract

A hybrid strategy for solar water splitting is exploited here based on a dye-sensitized photoelectrosynthesis cell (DSPEC) with a mesoporous SnO₂/TiO₂ core/shell nanostructured electrode derivatized with a surface-bound Ru(II) polypyridyl-based chromophore–catalyst assembly. The assembly, [(4,4’-(PO₃H₂)₂bpy)₂Ru(4-Mebpy-4’-bimpy)Ru(tpy)(OH₂)]⁴⁺ ([Ru_a^II-Ru_b^II-OH₂]⁴⁺, combines both a light absorber and a water oxidation catalyst in a single molecule. It was attached to the TiO₂ shell by phosphonate-surface oxide binding. The oxide-bound assembly was further stabilized on the surface by atomic layer deposition (ALD) of either Al₂O₃ or TiO₂ overlayers. Illumination of the resulting fluorine-doped tin oxide (FTO)|SnO₂/TiO₂|-[Ru_a^II-Ru_b^II-OH₂]⁴⁺(Al₂O₃ or TiO₂) photoanodes in photoelectrochemical cells with a Pt cathode and a small applied bias resulted in visible-light water splitting as shown by direct measurements of both evolved H₂ and O₂. The performance of the resulting DSPECs varies with shell thickness and the nature and extent of the oxide overlayer. Use of the SnO₂/TiO₂ core/shell compared with nanoITO/TiO₂ with the same assembly results in photocurrent enhancements of ∼5. Systematic variations in shell thickness and ALD overlayer lead to photocurrent densities as high as 1.97 mA/cm² with 445-nm, ∼90-mW/cm² illumination in a phosphate buffer at pH 7.

Although promising, significant challenges remain in the search for successful strategies for artificial photosynthesis by water splitting into oxygen and hydrogen or reduction of CO₂ to reduced forms of carbon (1–5). In a dye-sensitized photoelectrosynthesis cell (DSPEC), a wide band gap, nanoparticle oxide film, typically TiO₂, is derivatized with a surface-bound molecular assembly or assemblies for light absorption and catalysis (6–8). In a DSPEC, visible light is absorbed by a chromophore, initiating a series of events that culminate in water splitting: injection, intraassembly electron transfer, catalyst activation, and electron transfer to a cathode or photocathode for H₂ production. Sun and coworkers have recently demonstrated visible-light–driven water splitting with a coloading approach combining Ru(II) polypyridyl-based light absorbers and catalysts on TiO₂ (9). The efficiency of DSPEC devices is dependent on interfacial dynamics and competing kinetic processes. A major limiting factor is the requirement for accumulating multiple oxidative equivalents at a catalyst site to meet the 4e⁻/4H⁺ demands for oxidizing water to dioxygen (2H₂O - 4e⁻ - 4H⁺ → O₂) in competition with back electron transfer of injected electrons to the oxidized assembly.

One approach to achieving structural control of local electron transfer dynamics at the oxide interface in dye-sensitized devices is by use of nanostructured core/shell electrodes (10–12). In this approach, a mesoporous network of nanoparticles is uniformly coated with a thin oxide overlayer prepared by atomic layer deposition (ALD). We have used core/shell electrodes to demonstrate benzyl alcohol dehydrogenation (13). This approach has also been used to enhance the efficiency of dye-sensitized solar cells (14, 15). Recently, we described the use of a core/shell consisting of an inner core of a nanoparticle transparent conducting oxide, tin-doped indium oxide (nanoITO), and a thin outer shell of TiO₂ for water splitting by visible light (16). Derivatization of the nanoITO/TiO₂ core/shell electrode by surface binding of the chromophore–catalyst assembly, [(4,4’-(PO₃H₂)₂bpy)₂Ru_a(4-Mebpy-4’-bimpy)Ru_b(tpy)(OH₂)]⁴⁺ (1; -[Ru_a^II-Ru_b^II-OH₂]⁴⁺) shown in Fig. 1A, provided the basis for a photoanode in a DSPEC application with a Pt cathode for H₂ generation with a small applied bias in an acetate buffer at pH 4.6.

Fig. 1.

(A) Chemical structure of chromophore–catalyst assembly 1, -[Ru_a^II-Ru_b^II-OH₂]⁴⁺. (B) TEM depicting a core/shell nanostructure from 75 ALD cycles of TiO₂ deposited onto SnO₂ films on FTO glass (FTO|SnO₂/TiO₂(4.5 nm)|). (C) Cartoon depicting an ALD core/shell electrode surface with and without ALD overlayer stabilization of a surface-bound assembly.

Application of the core/shell structure led to a greatly enhanced efficiency for water splitting compared with mesoscopic, nanoparticle TiO₂ but the per-photon absorbed efficiency of the resulting DSPEC was relatively low and problems arose from long-term instability due to loss of the assembly from the oxide surface in the acetate buffer at pH 4.6. The latter is problematic because the rate of water oxidation is enhanced by added buffer bases, conditions that also enhance the rate of water oxidation (5, 17–24).

Here, we report a second-generation DSPEC based on a core/shell photoanode. It features both greatly enhanced efficiencies for visible-light–driven water splitting and stabilization of surface binding by the assembly. Enhanced efficiencies come from the use of a SnO₂ core in a SnO₂/TiO₂ core/shell structure. SnO₂ has a conduction band potential (E_CB) more positive than TiO₂ by ∼0.4 V. Once injection and electron transfer to the SnO₂ core has occurred, an internal potential gradient at the SnO₂/TiO₂ interface is established, inhibiting back electron transfer.

In the second-generation DSPEC, ALD is also used to stabilize oxide surface binding by the phosphonate-derivatized assembly. ALD deposition of overlayers of TiO₂ or Al₂O₃ has been shown to greatly enhance surface stability toward hydrolysis even in strongly basic solutions (25, 26). We show here, for assembly 1 surface-bound to SnO₂/TiO₂, that ALD overlayers of TiO₂ or Al₂O₃ provide both long-term stabilization on the oxide surface at pH 7 in a phosphate buffer, and, as a bonus, incrementally enhanced efficiencies for water splitting (23).

The underlying strategy behind the use of ALD for both core/shell structure and stabilized surface binding is illustrated in Fig. 1C. Detailed information about the mechanism and rate of water oxidation by the surface-bound assembly is available from a previous publication (27).

Results

Preparation of SnO₂/TiO₂ core/shell structures is described in SI Text. A transmission electron micrograph (TEM) is shown in Fig. 1B illustrating the core/shell structure prepared by uniformly coating a SnO₂ nanoparticle film with 75 ALD cycles of TiO₂.

Current–time (i–t) profiles were recorded at the photoanode of a photoelectrochemical cell at pH 4.6 with 20 mM acetate/acetic acid buffer or at pH 7 in a 0.1 M phosphate buffer with 0.5 M added LiClO₄. The DSPEC cell consisted of a fluorine-doped tin oxide (FTO)|SnO₂/TiO₂|-[Ru_a^II-Ru_b^II-OH₂]⁴⁺ core/shell photoanode with a Pt wire as the cathode. Illumination at 445 nm (FWHM 20 nm, ∼10 to ∼90 mW/cm², beam diameter 1 cm) was provided by a Lumencor SPECTRA seven-color solid-state light source.

Fig. 2A compares the results of short-term, current density–time DSPEC measurement at nanoITO/TiO₂ and SnO₂/TiO₂ core/shell electrodes with a nominal TiO₂ shell thickness of 3.3 with an applied bias of 200 mV vs. normal hydrogen electrode (NHE). As reported previously (16), cell performance is bias-dependent with an applied bias required to maximize photocurrent and H₂ evolution at the cathode.

Fig. 2.

(A) Photocurrent comparisons between SnO₂ and nanoITO core/TiO₂ photoanodes, FTO|SnO₂/TiO₂|-[Ru_a^II-Ru_b^II-OH₂]⁴⁺ (black) and FTO|nanoITO/TiO₂|-[Ru_a^II-Ru_b^II-OH₂]⁴⁺ (blue), with 50-cycle ALD TiO₂ shells (3.3 nm) derivatized with 1 with a Pt counterelectrode and 200 mV (vs. NHE) applied bias at pH 4.6 in 0.5 M LiClO₄ with 20 mM acetate/acetic acid buffer. The red trace shows the impact of a 10-cycle TiO₂ overlayer on the photocurrent output of the SnO₂ core/shell electrode. (B) Photocurrent–time curves for FTO|SnO₂/TiO₂(6.6 nm)|-[Ru_a^II-Ru_b^II-OH₂]⁴⁺ with 10 cycles of an added TiO₂ overlayer: 600 mV applied bias vs. NHE in 0.5 M LiClO₄ 20 mM in acetic acid/acetate buffer at pH 4.6 (red) and in a 0.1 M H₂PO₄⁻/HPO₄^2- buffer at pH 7 with the ionic strength adjusted to 0.5 M with NaClO₄ (black).

From the data in Fig. 2A and the data summary in Tables S1 and S2, a maximum initial photocurrent density of 0.48 mA/cm² was reached for the SnO₂/TiO₂ (3.3-nm) core/shell photoanode, falling to 0.1 mA/cm² after 10 s at the end of the initial current spike. The initial photocurrent increased to 0.79 mA/cm² with a 0.66-nm overlayer of TiO₂. The initial current spike arises from oxidation of the assembly to its steady-state form -[Ru_a^III-Ru_b^IV = O]⁴⁺ and from local capacitance effects (28, 29).

The small dark current at the end of the light-on/light-off cycles is a characteristic feature of DSPECs arising from electron equilibration by back electron transfer through the core/shell network to the partly oxidized, surface-bound assemblies.

Photocurrent comparisons were made after 10 s of 445-nm illumination at the end of the initial current spike at the onset of the plateau current. It is notable that in comparing nanoITO/TiO₂ and SnO₂/TiO₂ as core/shells under the same conditions, photocurrent increases of ∼5× are observed.

The results of a study of the effect of TiO₂ shell thickness on DSPEC performance for SnO₂/TiO₂ core/shells are summarized in Table 1. These experiments were conducted at pH 7 in a H₂PO₄⁻/HPO₄^2- buffer with [HPO₄^2-] ∼60 mM. Assembly-surface binding to the TiO₂ shell under these conditions was stabilized by ALD deposition of a 0.55-nm-thick overlayer of Al₂O₃.

Table 1.

TiO₂ shell thickness effect on current densities for FTO|SnO₂/TiO₂|-[Ru_a^II-Ru_b^II-OH₂]⁴⁺ as a function of TiO₂ shell thickness and 445-nm light intensity with 0.55 nm of Al₂O₃ overlayer

Light intensity at 445 nm (mW/cm2)	SnO2/TiO2(3.3 nm)|-[RuaII-RubII-OH2]4+(0.55 nm Al2O3)	SnO2/TiO2(4.5 nm)|-[RuaII-RubII-OH2]4+ (no overlayer)	SnO2/TiO2(4.5 nm)|-[RuaII-RubII-OH2]4+(0.55 nm Al2O3)	SnO2/TiO2(6.6 nm)|-[RuaII-RubII-OH2]4+(0.55 nm Al2O3)	SnO2|-[RuaII-RubII-OH2]4+(0.55 nm Al2O3)
15	0.80	0.93	1.39	1.14	0.04
56	1.04	1.20	1.89	1.78	0.06
86	1.02	1.26	1.97	1.77	0.02

Experiments were performed at room temperature in a H₂PO₄⁻/HPO₄^2- buffer ([HPO₄^2-] ∼60 mM) with the ionic strength adjusted to 0.5 with NaClO₄ and an external applied bias of 600 mV vs. NHE. The photocurrent densities in the table are reported in mA/cm².

The results in Table 1 show that compared with a SnO₂ core, the photocurrent density increases by greater than 30× for SnO₂/TiO₂ core/shells with shell thicknesses from 3.3 to 6.6 nm. The photocurrent density, which is dependent on TiO₂ shell thickness over the range 3.3, 4.5, 6.6 nm, is maximized at 4.5 nm (75 ALD cycles).

There is also a dependence on the number of ALD overlayer stabilization cycles and on the nature of the added overlayer. Based on photocurrent data at pH 4.6 and pH 7 in Table S3 and Fig. S1, photocurrent efficiencies for the assembly-based photoanode, FTO|SnO₂/TiO₂(6.6 nm)|-[Ru_a^II-Ru_b^II-OH₂]⁴⁺, were maximized by 0.33- or 0.55-nm-thick ALD overlayers of Al₂O₃ with enhancements greater for Al₂O₃ than for TiO₂. For the TiO₂ overlayers, the enhancement was greater for a TiO₂ overlayer of 0.6 nm compared with 1.2 nm. The highest photocurrents, 1.97 mA/cm², were reached with a 0.55-nm overlayer of Al₂O₃. Similar results were obtained for an electrode with an ∼4.5 nm TiO₂ shell as shown by the cyclic voltammetry measurements in the dark and under illumination in Fig. S2.

With no protective overlayer, at pH 4.6 in acetate buffer, loss of the assembly from the surface by hydrolysis is noticeable after a few minutes. At pH 7 in phosphate buffer, the loss is too rapid for current–time measurements. The assembly is stable on the surface under these conditions, however, with ALD-added overlayers of TiO₂ or Al₂O₃. The results of long-term photocurrent measurements on the ALD-stabilized photoanode, FTO|SnO₂/TiO₂(6.6 nm)|-[Ru_a^II-Ru_b^II-OH₂]⁴⁺-(0.6 nm)TiO₂, are shown in Fig. 2B. These results reveal an impressive enhancement in stability toward surface hydrolysis. A slow decrease in photocurrent is observed with time but it arises from instability toward ligand loss by the Ru (III) form of the chromophore in the assembly (30).

In previous experiments with assembly 1 on core/shell nanoITO/TiO₂, high faradaic efficiencies for H₂ production were reported with O₂ evolution confirmed by a rotating ring disk method (16). In the present study, H₂ and O₂ evolution from FTO|SnO₂/TiO₂(4.5 nm)|-[Ru_a^II-Ru_b^II-OH₂]⁴⁺(0.3 nm Al₂O₃) was confirmed by direct Clark-type oxygen and hydrogen microsensor (Unisense) measurements. The sensors, with tip diameters of 1.6 mm, were inserted into the DSPEC cell in the phosphate buffer at pH 7. A schematic is shown in Fig. S3. Current–time plots with 455-nm LED photolysis (46 mW/cm²) at an applied bias of 600 mV are shown in Fig. 3A and evolution curves for the appearance of H₂ and O₂ in Fig. 3B. In these experiments, a 600-mV bias was applied between FTO|SnO₂/TiO₂(4.5 nm)|-[Ru_a^II-Ru_b^II-OH₂]⁴⁺(0.3 nm Al₂O₃) photoanode and Pt counter in a two-electrode configuration. The faradaic efficiencies for H₂ and O₂ measured after 100 s of photolysis were 57% and 41%, respectively, which are low but typical for these measurements.

Fig. 3.

Photoelectrochemical water splitting by FTO|SnO₂/TiO₂(6.6nm)|-[Ru_a^II-Ru_b^II-OH₂]⁴⁺(0.3nmAl₂O₃) with a 600-mV applied bias in a 0.1 M H₂PO₄⁻/HPO₄^2- buffer at pH 7 at room temperature. The bias was applied across the working and counterelectrodes (the experiment was performed in a two-electrode configuration with the counter- and reference leads both connected to the Pt counterelectrode). The ionic strength was adjusted to 0.5 M with NaClO₄. Illumination was accomplished with a 455-nm LED at 46.2 mW/cm². (A) Photocurrent–time trace and (B) H₂ and O₂ evolution time traces recorded in concert with the photocurrent trace.

The results described here are a notable advance. Introduction of SnO₂/TiO₂ core/shells improves cell efficiencies by a factor of ∼5× (Table S1). Added ALD oxide overlayers stabilize surface binding over extended photolysis periods, even at pH 7 in phosphate buffers. Cell efficiencies can be manipulated by varying the core/shell material and its geometry. Under optimal conditions for a FTO|SnO₂/TiO₂(4.5 nm)|-[Ru_a^II-Ru_b^II-OH₂]⁴⁺(0.3 nm Al₂O₃) photoanode, a photocurrent density of 1.97 mA/cm² was reached for 445-nm water splitting. The underlying interfacial dynamics for the integrated molecular assembly–oxide device are currently under investigation by transient absorption and photocurrent measurements to assess the kinetic factors required to further increase cell efficiencies.

These results are important in expanding the scope of DSPEC water splitting by manipulating core/shell structure and incorporating ALD overlayer protection toward hydrolysis. Major challenges remain in maximizing solar light absorption, achieving higher levels of surface stabilization, and maximizing efficiencies, but the door appears to be open for a systematic exploitation of the DSPEC strategy.